Data Science and Applications for Modern Power Systems

Foreword
Preface
Acknowledgements
Contents
1 Data Perspective on Power Systems
1.1 What Is an Electric Grid?
1.2 A Data-Driven Perspective on Grid
1.2.1 Data-Driven Modeling and Monitoring
1.2.2 Data-Driven Control
1.2.3 Data-Driven Planning
1.3 Grid Data Availability
1.3.1 Outline of This Book
2 Basics of Power Systems
2.1 Participants
2.1.1 Generators
2.1.2 Prosumers
2.1.3 Aggregators
2.1.4 Utilities
2.1.5 System Operators
2.2 Flow of Power
2.3 Flow of Money
2.4 Flow of Information
2.4.1 Big Data Era
2.4.2 Challenges in Big Data Analytics
2.4.3 Look into the Future
3 Emerging Technology for Distributed Energy Resources
3.1 Distributed PV Generation
3.1.1 Introduction
3.1.2 Results
3.1.2.1 Scalable Deep Learning Model for Solar Panel Identification
3.1.2.2 Nationwide Solar Installation Database
3.1.2.3 Correlation Between Solar Deployment and Environmental/Socioeconomic Factors
3.1.2.4 Predictive Solar Deployment Model
3.1.3 Discussion
3.1.4 Experimental Procedures
3.1.4.1 Massive Satellite Imagery Dataset
3.1.4.2 System Detection Using Image Classification
3.1.4.3 Size Estimation Using Semi-supervised Segmentation
3.1.4.4 Distinguish Between Residential and Non-residential Solar
3.1.4.5 Predictive Solar Deployment Models
3.2 The Impact of Electric Vehicle Penetration
3.2.1 Introduction
3.2.2 The Impact of EV Charging Locations to the Power Grid
3.2.2.1 The Benefits of Problem Convexification
3.2.2.2 Sensitivity Analysis for the Optimization Variables
3.2.2.3 Sensitivity Analysis for Different Cost Components
3.2.3 The Impact of Choosing EV Routes for Charging
3.2.3.1 Numerical Results on Different EV Routes
3.2.3.2 Numerical Results on EV Numbers and Charging Time
3.2.4 Conclusion
4 Adapt Load Behavior as Technology Agnostic Solution
4.1 Consumer Segmentation
4.1.1 Introduction
4.1.1.1 Prior Work
4.1.2 Methodology
4.1.2.1 Total Daily Consumption Characterization
4.1.2.2 Encoding System Based on a Preprocessed Dictionary
4.1.2.3 Adaptive K-Means on Normalized Data
4.1.2.4 Hierarchical Clustering
4.1.3 Experiments on Data
4.1.3.1 Description of Smart Meter Data
4.1.3.2 Dictionary Generation on Real Usage Data
4.1.3.3 Dictionary Reduction via Hierarchical Clustering
4.1.3.4 Load Shape Analysis
4.1.4 Segmentation Analysis
4.1.4.1 Entropy Analysis
4.1.4.2 Shape Analysis
4.1.4.3 Multidimensional Segmentation
4.1.4.4 Spatial Locality Analysis
4.1.4.5 Temporal Locality Analysis
4.1.5 Impacts on Load Forecasting
4.1.6 Conclusion and Future Work
4.2 Consumer Targeting
4.2.1 Introduction
4.2.2 Methodology
4.2.2.1 Maximizing Demand Response Reliability
4.2.2.2 Response Modeling
4.2.3 Algorithm
4.2.3.1 Optimization Problem Transformation
4.2.3.2 Previous Approaches to Solve the SKP
4.2.3.3 Stochastic Knapsack Problem-Solving
4.2.4 Experiment on Data
4.2.4.1 Description of Data
4.2.4.2 Consumption Model Fitting Result
4.2.4.3 Targeting Result Analysis
4.2.5 Conclusion
4.3 Demand Response
4.3.1 Introduction
4.3.2 Probabilistic Baseline Estimation for Residential Customers
4.3.3 Probabilistic Baseline Estimation via Gaussian Process Regression
4.3.4 Feature Extraction: Covariance Function Design
4.3.4.1 Embedding Distance-Based Correlation
4.3.4.2 Embedding Periodic Pattern
4.3.4.3 Embedding Piecewise Linear Pattern in Temperature
4.3.4.4 Embedding More Functions
4.3.5 Utilizing Probabilistic Estimate for Fair Payment to Residential Customers
4.3.6 Simulation Result
4.3.6.1 Improved Daily Accuracy Without Day Aggregation
4.3.6.2 Reduced Relative Confidence Intervals with Day Aggregation
4.3.6.3 Reduced Relative Error with Day Aggregation
4.3.6.4 Computational Time
4.3.7 Conclusion
4.4 Energy Coupon as Demand Response
4.4.1 Introduction
4.4.2 System Overview
4.4.3 Experimental Algorithms
4.4.3.1 Price Prediction
4.4.3.2 Baseline Estimate
4.4.3.3 Individualized Target Setting and Coupon Generation
4.4.3.4 Lottery Algorithms
4.4.4 Experimental Design
4.4.4.1 Brief Summary of Experiment ('16)
4.4.4.2 Subject in Experiment ('17)
4.4.4.3 Procedure in Experiment ('17)
4.4.5 Results and Discussion
4.4.5.1 Energy-Saving for the Treatment Group
4.4.5.2 Comparison Between Active and Inactive Subjects in Treatment Group
4.4.5.3 Comparison Between Subjects in Treatment Group Facing Fixed/Dynamic Coupons
4.4.5.4 Financial Benefit Analysis
4.4.5.5 Influence of the Lottery on Human Behavior
4.4.5.6 Comparison with Previous CPP Experiment
4.4.5.7 Cost-Saving Decomposition
4.4.6 Conclusion
5 Use of Energy Storage as a Means of Managing Variability
5.1 Adding Storage to the Mix
5.1.1 Introduction
5.1.2 Formulation
5.1.2.1 Solar Generation
5.1.2.2 Load
5.1.2.3 Storage
5.1.2.4 Reliability Value
5.1.3 Optimal Investment Problem
5.1.4 Main Results
5.1.4.1 Reliability Value and Optimal Investment Decision
5.1.4.2 Example: Deterministic Case
5.1.5 Case Studies
5.1.5.1 A Benchmark Model
5.1.5.2 Data Description
5.1.5.3 Theoretical Estimates
5.1.5.4 Realistic Results
5.1.5.5 Optimal Investment Decision
5.1.5.6 Discussions
5.1.6 Conclusion
5.2 Long-Term Planning via Scenario Approach
5.2.1 Introduction
5.2.2 Probabilistic Storage Planning
5.2.2.1 Deterministic Storage Planning
5.2.2.2 Storage Planning with Probabilistic Guarantees
5.2.2.3 Structure of the Storage Planning Problem
5.2.3 Solving Probabilistic Storage Planning
5.2.3.1 Introduction to the Scenario Approach
5.2.3.2 Solving Probabilistic Storage Planning via the Scenario Approach
5.2.3.3 Sub-gradient Cutting-Plane Method
5.2.4 Case Study
5.2.4.1 Settings
5.2.4.2 Numerical Results
5.2.4.3 Discussions
5.2.5 Conclusion
5.3 Utility's Procurement with Storage and/or Demand Response
5.3.1 Introduction
5.3.2 Problem Formulation
5.3.2.1 System Model
5.3.2.2 Optimization Problem
5.3.2.3 Model-Based Solution for Benchmarking
5.3.2.4 Problem Statement
5.3.3 Model-Free Privacy-Preserving Optimization and Control Framework
5.3.3.1 Stage 1: Optimization
5.3.3.2 Stage 2: Private Control Implementation
5.3.4 Case Study
5.3.5 Conclusion and Future Work
6 Forecast for the Future
6.1 Forecasting
6.1.1 Introduction
6.1.2 Theoretical Analysis
6.1.2.1 Notations
6.1.2.2 Security-Constrained Economic Dispatch
6.1.2.3 SCED Analysis via MLP
6.1.2.4 An Illustrative Example
6.1.3 SPRs with Varying Parameters
6.1.3.1 Dynamic Line Rating
6.1.3.2 Ramping Constraints
6.1.4 A Data-Driven Approach to Identifying SPRs
6.1.4.1 The SPR Identification Problem
6.1.4.2 A Data-Driven Approach
6.1.5 Case Study
6.1.5.1 Performance Metrics
6.1.5.2 Static SCED with Static Line Ratings
6.1.5.3 Static SCED with Dynamic Line Ratings
6.1.5.4 Case Studies with Ramp Constraints
6.1.6 The Impact of Nodal Load Information
6.1.6.1 On Nodal Load Levels
6.1.6.2 Incomplete Load Information
6.1.7 Discussions
6.1.7.1 On Posterior Probabilities
6.1.7.2 On the Computational Cost
6.1.7.3 On Generation Offer Prices
6.1.7.4 LMPs with Loss Components
6.1.8 Conclusions
6.2 Price Prediction
6.2.1 Introduction
6.2.2 Problem Formulation
6.2.2.1 Direct Method (Price-to-Price Method)
6.2.2.2 Rerouted Method (Two-Stage Method)
6.2.3 Machine Learning Methods
6.2.3.1 Overview of Methods
6.2.3.2 Performance Evaluation Metric
6.2.4 Numerical Results
6.2.4.1 Data Preparation
6.2.4.2 Benchmark
6.2.5 Conclusion
6.3 Residential Appliances
6.3.1 Introduction
6.3.1.1 Related Work
6.3.1.2 Summary of Contributions
6.3.2 Appliance Load Characterization
6.3.2.1 Discrete Operating States
6.3.2.2 Duration Analysis
6.3.3 Hidden Semi-Markov Model
6.3.4 Appliance Load Model
6.3.4.1 Conditional HSMM
6.3.4.2 Parameter Estimation
6.3.4.3 State-Specific Model
6.3.4.4 Weighted Logistic Regression
6.3.5 Short-Term Load Forecasting
6.3.6 Case Studies
6.3.6.1 Data
6.3.6.2 Parameter Specification
6.3.6.3 Performance Metric
6.3.6.4 Load Forecasting for Individual Appliances
6.3.6.5 Load Aggregation and Model Refinements for A/Cs
6.3.6.6 Scalability and Performance
6.3.7 Conclusion
7 Design New Markets
7.1 Scenario-Based Stochastic Dispatch
7.1.1 Introduction
7.1.2 Taxonomy of Look-Ahead Economic Dispatch Under Uncertainty
7.1.2.1 Deterministic, Stochastic and Robust LAED
7.1.2.2 Scenario Approach LAED
7.1.3 Computational Algorithm to Solve the Scenario Approach Economic Dispatch
7.1.3.1 The A Priori Scenario Approach Method
7.1.3.2 Sampling and Discarding Approach in Sc-LAED
7.1.3.3 The A Posteriori Scenario Approach Method
7.1.4 Case Study
7.1.4.1 Extreme Ramping Test: Scenario vs. Deterministic and Robust LAED
7.1.4.2 Risk and Complexity: Considering All Constraints in the Sc-LAED
7.1.5 Conclusion
7.2 ISO Dispatch
7.2.1 Introduction
7.2.2 Problem Formulation
7.2.2.1 DRP as a Supplier in Day-Ahead Market
7.2.2.2 Decision Curve for DRP
7.2.2.3 Uncertainty of DR
7.2.3 Economic Dispatch Methods in Day-Ahead Market
7.2.3.1 Deterministic Model
7.2.3.2 Stochastic Model
7.2.3.3 Robust Model
7.2.3.4 Scenario Approach Model
7.2.3.5 Realization Cost
7.2.4 Numerical Examples
7.2.4.1 3-Bus System with One DRP
7.2.4.2 Simulation Results for Economic Dispatch
7.2.4.3 Trade-Off Between Feasibility and Performance
7.2.4.4 Influence of δ on DR Acceptance
7.2.4.5 IEEE 14-Bus System with Two DRPs
7.2.5 Conclusion
8 Streaming Monitoring and Control for Real-Time Grid Operation
8.1 Learning the Network
8.1.1 Introduction
8.1.2 Probabilistic Modeling of Network Voltages via Graphical Modeling
8.1.2.1 Problem Definition
8.1.3 Mutual Information-Based Algorithm for Distribution Grids
8.1.3.1 Why Mutual Information-Based Algorithm Works?
8.1.3.2 Adaptation for Distribution Grid with a Loop
8.1.3.3 Adaptation for Smart Meter with Voltage Magnitude Data
8.1.3.4 Limitations of the Method
8.1.4 Simulations
8.1.4.1 Tree Networks without DERs
8.1.4.2 Tree Networks with DERs
8.1.4.3 Networks with a Loop
8.1.4.4 Algorithm Sensitivities
8.1.5 Conclusion
8.2 State Estimation of the Steady-State
8.2.1 Introduction
8.2.2 Graphical Modeling
8.2.3 Distributed Joint State Estimation
8.2.3.1 An Objective Prior Probability
8.2.3.2 Embedding Physical Laws in the Conditional Probability
8.2.3.3 Marginalization for Interested State in Tree-Structured Networks
8.2.3.4 From Tree Structure for Distribution Grids to Mesh Structure for Transmission Grids
8.2.3.5 Improvement over Convergence, Optimality, and Memory Requirement
8.2.3.6 Algorithm Summary
8.2.4 Illustration Using an Example
8.2.5 Numerical Results
8.2.6 Error Domain Comparison Based on Mean Estimate
8.2.7 Variance Estimate
8.2.8 Computational Cost
8.2.8.1 Improvement over Convergence, Optimality, and Memory
8.2.8.2 The Impact of PMU Measurements
8.2.9 Conclusion and Future Research
8.3 Voltage Regulation Based on RL
8.3.1 Introduction
8.3.2 Preliminaries
8.3.3 Markov Decision Process and Reinforcement Learning
8.3.4 Voltage Regulation as an RL Problem
8.3.4.1 State Space
8.3.4.2 Action Space
8.3.4.3 Transition Model
8.3.4.4 Reward Function
8.3.5 Control Policy Architecture and Optimization
8.3.6 Numerical Simulation
8.3.6.1 Simulation Setup
8.3.6.2 Case Study on a Smaller (16-bus) Subsystem
8.3.6.3 Case Study on a Larger (194-bus) Subsystem
8.3.7 Conclusion
9 Using PMU Data for Anomaly Detection and Localization
9.1 Dynamics from PMU
9.1.1 Introduction
9.1.2 Linear Analysis of Synchrophasor Dimensionality
9.1.3 Online Event Detection Using PMU Data
9.1.3.1 Adaptive Training
9.1.3.2 Robust Online Monitoring
9.1.4 Numerical Examples
9.1.4.1 Dimensionality Reduction of Synchrophasor Data
9.1.4.2 Dimensionality Reduction of Realistic Texas Data
9.1.4.3 Online Event Detection Using the Early Event Detection Algorithm
9.1.5 Conclusion
9.2 Asset Management
9.2.1 Introduction
9.2.2 Localization of Forced Oscillations and Challenges
9.2.2.1 Mathematical Interpretation
9.2.2.2 Main Challenges of Pinpointing the Sources of Forced Oscillation
9.2.3 Problem Formulation and Proposed Methodology
9.2.3.1 Problem Formulation
9.2.3.2 FO Localization Algorithm for Real-Time Operation
9.2.4 Theoretical Interpretation of the RPCA-Based Algorithm
9.2.4.1 PMU Measurement Decomposition
9.2.4.2 Observations on the Resonance Component and the Resonance-Free Component
9.2.4.3 Low-Rank Nature of Resonance Component Matrix
9.2.5 Case Study
9.2.5.1 Performance Evaluation of the Localization Algorithms in Benchmark Systems
9.2.5.2 Algorithm Robustness
9.2.5.3 Impact of Noise on Algorithm Performance
9.2.5.4 Comparison with Energy-Based Localization Method
9.2.6 Conclusion
References
Index